Update readme.

Finished first half of read me,

README.md

@@ -1,3 +1,60 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+---
+
+
+<p align="center">
+    
+
+<img src="https://cdn-uploads.huggingface.co/production/uploads/638aeb590f10aa3064f42a80/fwiBrNZ0XEU9LouwbbFga.png" width="600"/>
+
+<p>
+<p align="center">
+           <a href="https://civitai.com/user/lilylilith"><b>Civitai</b></a>&nbsp&nbsp | &nbsp&nbsp☕️ <a href="https://ko-fi.com/lilllithhhhh">Ko-fi (support my work)</a>&nbsp&nbsp
+
+<img src="https://cdn-uploads.huggingface.co/production/uploads/638aeb590f10aa3064f42a80/ZMr69PCuSIKuCyeMJ0TS2.png" width="1600"/>
+
+# Introduction      
+
+Creating poses for qwen image edit (2511) or its variants can be difficult. Even with openpose being baked into the model, it can still look like something is off with the result, such as the depth being incorrect or the pose not fully matching the character, with distortions that you wouldn't expect. Additionally, creating them by hand can take a bit too long, especially if you are going the blender openpose rig route. What if instead you could just skip openpose entirely and write a simple (but albeit lengthy) prompt to copy the pose of any image on your computer?
+
+Well, this is what the AnyPose LoRAs sets out to do. With just a single reference image as a pose guide, you can pilot any image to follow that pose. No control net needed. 
+
+# Showcase
+
+AnyPose works with simple poses, like T or A poses,
+
+
+![tposeExample](https://cdn-uploads.huggingface.co/production/uploads/638aeb590f10aa3064f42a80/21W48KegxqP7JvMTPWg9W.png#center)
+
+or even complex poses like yoga poses:
+
+
+![yogaExample](https://cdn-uploads.huggingface.co/production/uploads/638aeb590f10aa3064f42a80/5gkuKBKxTe3NUf_t7_6PI.png#center)
+
+
+# Important Notes:
+
+Now there are some notes that you must know about before using this. The creation of the dataset for LoRA sounds very intuitive, right? Just pose two characters in the same pose (I used Blender to pose dense 3d characters in various precocious poses that I knew Qwen Edit wouldn't understand ((think handstands, bridges, bends, etc)); so this was easy to do. 
+
+However it was very hard to decide the method on how the final processed image should look like. Should the final output be the same scene, but with the new pose? Or should the final composite be instead with the character "swapped in" with the other character, in the second image's scene? The first one sounds the most intuitive, yet it comes with some drawbacks such as hallucinated backdrops inpainted in. Yet the second option makes it a lot less useful if you just wanted to change the pose of the person in the current scene. It's just very unintuitive. So I tried to do both.
+
+In retrospect I don't think it was the best choice, as there are times where the person replaces the person doing the pose in the second image, inheriting its background; or elements of the previous scene crop into the current one. Sometimes, though it does keep the same scene. This doesn't sound good for a consistency point of view, however it existing brings major advantages, such as it will always fill in the missing areas.
+
+For example, if you want to pose a character who's full body is not shown, such as the lower half of their body, AnyPose will automatically fill in the 'unknown' areas of the image with a blend of the original image:
+
+
+In order to fix this, you can append a prompt at the end of the trigger prompt in order to add or remove elements from the output. For example, for the unknown areas, you can say something like "Wearing white leggings." in order to tell the model what she is wearing:
+
+
+(However, let it be known that the best way to use AnyPose is to have the character's full body shown (and or full environment shown) in the initial image, so that the resulting pose becomes consistent and qwen edit doesn't have to guess.)
+
+
+Rest: TBD. Finish soon.
+
+
+
+
+
+
+